Mailbox Zero: testing a junk-mail startup in 3 weeks
Why is there no unsubscribe button for paper mail? I wanted to know whether people would pay for one, so I built Mailbox Zero and put real money behind a test. Here's how it went, including the part that didn't work.
The problem
Our house was a rental before we bought it, so along with our own junk mail we get a steady stream for people who don't live here anymore. Email has had a legally required way to unsubscribe for two decades. Paper mail has nothing like it (and the Postal Service is paid by the senders, so don't hold your breath).
Since August I've run a homemade fix on our own mail. One agent reads the free USPS Informed Delivery email each morning and tells me whether anything is worth the walk. Another tracks opt-out requests to each sender and checks 8 weeks later whether the mail actually stopped. Six weeks in, 10 senders had gone quiet.
The real question was whether strangers would pay for it.
What I built
getmailboxzero.com went live two days after I started. Over the next three weeks it grew to 50+ pages:
- Step-by-step opt-out guides for 25+ companies, each route checked on the company's own site. No phone numbers, ever. Nobody wants to sit on hold to stop a catalog.
- Guides for mail addressed to former residents, and for new homeowners buried in "welcome" offers.
- A done-for-you offer: send photos of your junk mail and we file every opt-out request for you. You approve each one first.
I built it with Claude as my build team, and Claude agents ran the back office: a Monday report on ads, search traffic and sign-ups, a Thursday agent that drafted new pages for my review, an hourly inbox agent that drafted every customer reply, and a scanner for reporters' source requests. Every agent was draft-only, so nothing got published, sent or spent without my OK.
How I tested it
I split the market into seven customer types (former residents' mail, new homeowners, specific companies, credit card offers, aging parents, small businesses and a general control) and gave each one its own Google Ads campaign and landing pages. Budget: $275, later stretched by a week to test new ads.
Before launch I wrote down how I'd judge it: search demand, click-through rate and cost per click first, with sign-ups only as a tiebreaker, since a few hundred dollars buys too few clicks to trust a sign-up rate. I also had a second AI model review the site, the ads and the plan on its own, so I wasn't grading my own homework.
What happened
| Search ads Oct. 31 to Nov. 16 | 495 impressions, 73 clicks, a 14.75% click-through rate, more than double the 6.66% all-industry average*, at $3.57 a click. |
|---|---|
| Google search free traffic | Page one for several "stop [company] mail" searches within days of launch, and about 1,500 impressions a week by mid-November. Clicks stayed small, about 6 a week. |
| Sign-ups from strangers | Zero, across three versions of the offer (priced and free), from roughly 85 to 100 visitors from search. |
*WordStream and LocaliQ, 2025 Google Ads Benchmarks (16,000+ U.S. search campaigns).
What I learned
People want the fix, not a service
The most common click on the site was out to OptOutPrescreen.com, the free official opt-out for credit card offers. They got their answer in five minutes and left. Great for them. Less great for a business model.
Selling the service cut clicks in half
Ads offering done-for-you help got about half the click-through rate of ads offering the free answer (12% vs. 27%).
Free didn't change the zero
Priced and free versions both got zero, so price wasn't the blocker. The likely buyers, like a new owner of a former rental or a family sorting a parent's mail, probably need a warmer channel than search ads. Realtors and property managers are the obvious next one to test.
The prize is small
PaperKarma, the best-known paid app, charges $24.99 a year. Stopping junk mail is mostly a one-time job, so a subscription is a tough sell.
Where it stands
I parked Mailbox Zero on November 16. The free guides stay up for about $9 a month, and an agent still sends me a weekly report. I'll reopen it if a stranger signs up, a reporter bites or search traffic grows past a set threshold.
Three weeks and under $350 bought a clear answer before any money went into building the full product.